5G MetaAAU with Extremely Large Antenna Array

5G MetaAAU with Extremely Large Antenna Array

Noaman Tauq

Saudi Telecom Company

AbstractThe rapid densication of 5G New Radio (NR) networks has intensied the need for antenna systems that simultaneously improve coverage, capacity, and energy efciency without increasing site-footprint. This paper presents a eld evaluation of Huaweis 5G MetaAAU an Active Antenna Unit built around an Extremely Large Antenna Array (ELAA) of 384 dipole elements, double the count of a conventional 64T64R AAU deployed on the 3.63.8 GHz mid-band (n78) in a 5G NR network. The MetaAAU integrates the Signal Direct Injection Feeding (SDIF) architecture and an ultra-light integrated array to achieve a 3 dB gain in both uplink (UL) and downlink (DL) coverage. Coupled with the proprietary Adaptive High- Resolution (AHR) Turbo beamforming algorithm, the system delivers precise, dynamic, and targeted beam management across dense user scenarios. Trial results from two live commercial macro sites conrm 1820% extension of cell radius, 3040% improvement in cell-edge throughput, 1030% gain in indoor user experience, and 1020% increase in total cell trafc, while sustaining approximately 30% reduction in energy consumption compared with conventional AAU deployments. These ndings position MetaAAU as a high-impact upgrade path for operators pursuing simultaneous network performance uplift and carbon footprint reduction under 5G densication strategies.

Index TermsArray (ELAA), Massive MIMO, AHR Turbo beamforming, n78 mid-band, coverage extension, energy ef- ciency, Signal Direct Injection Feeding (SDIF).

1. Introduction

   The deployment of fth-generation (5G) New Radio (NR) networks has introduced stringent requirements on antenna systems in terms of spectral efciency, coverage, user through- put, and energy consumption [1].Massive Multiple-Input Multiple-Output (mMIMO) technology, based onactive an- tenna units (AAU) with large horizontal and vertical antenna arrays, has become a cornerstone of 5G RAN design on mid- band spectrum [2]. Conventional 64T64R AAUs operating in the 3.5 GHz band typically employ 192 radiating dipole elements arranged in a 12×8×2 conguration, providing three- dimensional (3D) beamforming through digital pre-coding [3]. Despite proven performance gains over 4G macro radios, con- ventional mMIMO AAUs face fundamental limits in cell-edge coverage, deep indoor penetration, and co-channel interference management especially as 5G subscriptions grow and inter-site distance shrinks [4]. These constraints have spurred research into Extremely Large Antenna Arrays (ELAA), which extend the aperture of the array beyond conventional bounds to exploit near-eld propagation effects, higher spatial resolution, and improved spatial multiplexing gain [5]. Huawei Tech- nologies introduced MetaAAU as a commercial realisation of ELAA for the sub-6 GHz mid-band, doubling the antenna count to 384 dipole elements within a physically integrated

   enclosure. The product combines hardware innovations namely the Signal Direct Injection Feeding (SDIF) architecture and an ultra-light integrated array with an advanced Adap- tive High-Resolution (AHR) Turbo beamforming algorithm to translate the larger aperture into tangible network KPI improvements. This paper makes the following contributions: leftmargin=*, label=1.

   1. A detailed description of MetaAAU hardware architec- ture, contrasting ELAA-based design with conventional 64T64R AAU.

   2. Analysis of the AHR Turbo algorithm and its role in precision beam management.

   3. Empirical trial results from two live commercial 5G sites covering coverage, throughput, user count, and trafc metrics.

   4. Discussion of energy efciency outcomes and deploy- ment implications for network operators.

   The remainder of the paper is structured as follows. Sec- tion II reviews related work on mMIMO and ELAA systems. Section III describes the MetaAAU system architecture. Sec- tion IV presents the trial methodology and site congurations. Section V discusses measured KPI results. Section VI ad- dresses energy efciency aspects. Section VII concludes the paper.

2. BACKGROUND AND RELATED WORK

   Practical 5G NR deployments conrmed theoretical predic- tions: eld trials with 32T32R and 64T64R systems showed 35 dB coverage improvement and 24× capacity gain over LTE MIMO in mid-band spectrum [8]. However, these de- ployments revealed that spatial multiplexing gain saturates beyond a certain antenna count when array aperture is not commensurately scaled motivating exploration of ELAA architectures [9].

   ELAA research has highlighted two distinct propagation regimes. In the far-eld (Fraunhofer) region, conventional planar wavefront assumptions hold and standard digital pre- coding strategies remain applicable. In the near-eld (Fresnel) region, the wavefront curvature must be explicitly modelled; beam focusing on specic spatial locations rather than angular directions becomes possible, enabling simultaneous coverage of multiple users at the same angle but different distances [10].

   On the hardware side, the feeding network architecture is critical for large-aperture arrays. Traditional distributed feeding introduces signicant power losses and inter-element coupling at n78 frequencies. Signal Direct Injection Feeding (SDIF), where each transceiver chain drives its radiating

   element directly without a passive distribution network, elimi- nates these losses and improves array efciency at the cost of increased transceiver count [11].

   Algorithmic innovations have paralleled hardware ad- vances. Hybrid beamforming architectures combine a reduced- dimension digital layer with an analogue phase-shift network to reduce hardware complexity while approaching fully-digital performance [12]. More recently, machine-learning-assisted beam management including beam prediction, tracking, and codebook optimization has been shown to reduce beam management overhead and improve beam accuracy under mobility. The AHR Turbo algorithm evaluated in this paper belongs to this category of adaptive, data-driven beam management.

3. SYSTEM ARCHITECTURE AND FEATURE DESCRIPTION

   64T64R AAU, the CFN introduces insertion losses of 1.52 dB per feeding path, directly reducing the effective power radiated by each sub-array. At doubled element count, the feeding loss problem would be compounding if a conventional CFN architecture were retained.

   MetaAAU employs Signal Direct Injection Feeding (SDIF), which eliminates the intermediate passive distribution network by connecting each transceiver output directly to its dedi- cated radiating sub-array [11]. This point-to-point injection approach: (i) removes CFN insertion loss from the power bud- get, improving radiated efciency; (ii) reduces inter-element coupling that degrades array radiation pattern delity; and

   (iii) allows more precise per-element amplitude and phase control, enabling ner beam resolution. Together with the ULIA mechanical integration, SDIF allows the 384-element array to maintain power efciency parity with a 192-element conventional design.

   1. Hardware Design

      The MetaAAU operates in the 3.63.8 GHz frequency band (3GPP NR band n78) with a 200 MHz channel bandwidth and a 64-transceiver (64T64R) digital front-end identical in radio-frequency (RF) channel count to conventional AAU counterparts but radically different in antenna realsation.

      The fundamental hardware innovation lies in the antenna array. Conventional 64T64R AAUs employ 192 radiating dipole elements arranged as 2 cross-polarised dipoles per sub- array across 96 sub-array positions, yielding a compact form

      factor of approximately 730 mm × 395 mm × 180 mm and a

      weight below 28 kg. MetaAAU doubles the radiating aperture to 384 dipole elements by extending the vertical aperture to 1450 mm while maintaining the horizontal dimension at 400 mm, resulting in a weight below 30 kg a remarkably modest mass increase for doubled antenna count, achieved through the ultra-light integrated array (ULIA) mechanical design.

      TABLE I

      Hardware Specification Comparison: MetaAAU vs. Conventional AAU

      Parameter MetaAAU Conventional AAU

      Frequency Band	36003800 MHz	36003800 MHz
      Output Power	320 W	320 W
      TRx Channels	64T64R	64T64R
      Bandwidth	200 MHz	200 MHz
      Antenna Elements	384 dipoles	192 dipoles
      Dimensions (mm)	1450×400×180	730×395×180
      Weight	<30 kg	<28 kg

      The doubled antenna aperture directly increases the effective isotropic radiated power (EIRP) and, more importantly, the array gain in the elevation domain providing the mechanism for 3 dB UL/DL coverage improvement reported in eld trials.

   2. Signal Direct Injection Feeding (SDIF)

      Conventional large-scale arrays distribute the RF signal from each transceiver output through a passive Corporate Feed Network (CFN) before reaching the radiating elements. For a

   3. Adaptive High-Resolution (AHR) Turbo Beamforming

      The hardware aperture advantage of MetaAAU is fully exploited only when the beamforming algorithm can trans- late additional degrees of freedom into measurable user-level improvements. The Adaptive High-Resolution (AHR) Turbo algorithm, deployed as a software layer on the baseband unit, addresses this requirement through three coordinated mechanisms:

      1. Precision channel estimation: AHR Turbo exploits the enlarged spatial aperture to apply high-resolution angle-of-arrival (AoA) and angle-of-departure (AoD) estimation, signicantly reducing the angular uncertainty of user location compared with conventional DFT-based codebooks calibrated for 192-element arrays.

      2. Dynamic beam adaptation: Rather than selecting from a xed codebook, the algorithm computes user-specic beamforming weights in the time domain using a near- real-time optimisation loop, adapting to rapid channel variations caused by user mobility, multipath dynamics, and changing inter-cell interference conditions.

      3. Targeted spatial multiplexing: The improved angular resolution allows the gNB to spatially separate users who would be indistinguishable to a narrower array, increasing the effective number of simultaneously served spatial layers contributing to both cell-average and cell-edge throughput gains.

   Together, these capabilities make AHR Turbo the primary software enabler for MetaAAUs performance gains, particu- larly in dense urban scenarios where interference management and coverage at cell edges are the binding constraints.

4. Methodology

   A. Sites and Conguration

   The eld evaluation was conducted in March 2023 across two live commercial 5G macro sites representing distinct deployment environments:

   • Site A Urban macro site with high average user count and dense trafc load.

   • Site B Suburban macro site with moderate user count and mixed indoor/outdoor coverage requirements.

   At each site, the conventional 64T64R AAU was replaced with MetaAAU operating in identical RF conguration: band n78 (36003800 MHz), 200 MHz bandwidth, 320 W output power, 64T64R transceivers. The AHR Turbo beamforming algorithm was activated on both trial sectors. All other RAN parameter settings scheduling policies, inter-frequency han- dover thresholds, power control parameters were kept constant throughout the trial to isolate the hardware and algorithm contribution.

   Indoor penetration improvement of 1030% in received signal level was also recorded, directly attributable to the 3 dB link budget uplift overcoming O2I penetration loss. Cell-edge (5th-percentile) RSRP improved by 3 dB, with 30% better cell- edge user experience in terms of RSRP and SINR distributions.

   B. Throughput and Capacity Improvements

   The AHR Turbo algorithms spatial multiplexing gains translate directly into throughput improvements as summarised below:

   B. Key Performance Indicators (KPIs)

   The following KPIs were collected before (baseline) and after (trial) MetaAAU activation from the Operations and Maintenance (OAM) system and drive-test measurement logs:

   • Average Access Timing Advance (TA) proxy for average cell radius in the uplink.

   • Average UE count per cell (active users scheduled).

   • Total cell DL/UL trafc volume (GBytes per hour).

   • DL/UL reference signal received power (RSRP) at the cell edge (5tp0th percentile CDF).

   • Cell-average and cell-edge DL/UL throughput (Mbps).

   • DL/UL Indoor signal strength (penetration loss measure- ment).

5. Results

   A. Coverage Extension

   The most fundamental impact of the doubled antenna aper- ture is a 3 dB gain in both UL and DL link budget. Using the standard outdoor-to-indoor (O2I) loss model for mid-band frequencies, a 3 dB link budget improvement translates to an 1820% extension of cell radius:

   r = r0 h(10L/10nPL ) 1i × 100% (1)

   TABLE III

   Throughput KPI Improvements MetaAAU vs. Baseline AAU

   KPI Improvement (%)

   Average DL/UL throughput (cell-average) +20 to +30 Cell-edge DL/UL throughput +30 to +40

   Indoor user experience +10 to +30 Cell-edge user experience (RSRP/SINR) +30

   The larger relative gain at the cell edge (+3040%) com- pared with the cell average (+2030%) is a characteristic signature of improved beam precision: cell-edge users, which previously suffered from weaker, less accurate beams and higher inter-cell interference leakage, benet disproportion- ately from both the 3 dB aperture gain and the tighter spatial nulling of AHR Turbo.

   C. Trafc and User Count

   Network-level trafc and user statistics reect the combined effect of expanded coverage area and improved spectral ef- ciency per user:

   TABLE IV

   Traffic and User Count Results per Trial Site

   KPI Site A (Urban) Site B (Suburban)

   Avg. User Count Increase	+38.41%	+14.79%
   Total Trafc Increase	+21.48%	+10.03%

   where L = 3 dB is the additional link budget, nPL is the

   path-loss exponent (typically 3.54.0 for urban mid-band), and r0 is the baseline cell radius. For nPL = 3.76 (3GPP UMa model for n78), equation (1) yields r 18.4%, consistent with the 1820% measured values.

   Average Access TA measurements from the two trial sites conrm this coverage extension quantitatively:

   TABLE II

   Coverage Extension Average Access TA Improvement

   Site Baseline TA (m) Trial TA (m) Gain (m)

   Site A +124.8

   Site B +125.97

   TA values converted to distance using TA step = 16Ts; baseline absolute values not disclosed. The consistent 125 m TA increase across both sites, despite differing propagation environments, indicates that the coverage gain is primarily driven by the MetaAAU hardware contribution rather than site- specic factors.

   The higher user count gain at Site A (+38.4%) versus Site B (+14.8%) reects the denser urbanisation of the Site A: more potential users reside within the extended coverage footprint, and the improved cell-edge SINR enables the scheduler to admit and sustain more concurrent UEs. The trafc increase, at +1020% across both sites, is consistent with the +1020% range projected from capacity modelling of the AHR Turbo spatial multiplexing improvement.

   Overall, the consolidated trial outcomes across both sites conrm:

   leftmargin=*

   +3 dB UL/DL coverage with 1820% cell radius exten- sion.

   +1030% indoor user experience improvement.

   +30% cell-edge user experience improvement.

   +2030% average throughput improvement (DL and UL).

   +3040% cell-edge throughput improvement (DL and UL).

   +1020% total cell trafc increase.

   +1540% active user count increase.

6. Energy Efficiency

   A critical consideration for large-antenna deployments is power consumption. Doubling the antenna count from 192 to 384 elements naively doubles the number of RF chains; with- out countermeasures, this would increase site power consump- tion proportionally and negate any green-network benets.

   MetaAAU addresses this through three mechanisms:

   1. SDIF efciency gain: Elimination of CFN insertion loss (1.52 dB per path) reduces per-element PA drive requirement, partially offsetting the increased transceiver count.

   2. AHR Turbo coverage-per-watt improvement: Tighter, more accurate beams reduce wasted radiated power in directions without active users, effectively improving the useful-signal fraction of total radiated power.

   3. Intelligent sleep/muting: AHR Turbo integrates intel- ligent carrier and channel element muting during low- trafc periods, reducing baseband and RF power when the spatial multiplexing capability is not required.

   The net outcome of these mechanisms is an approximately 30% reduction in energy consumption per bit compared with a conventional AAU baseline at equivalent trafc load positioning MetaAAU as a viable energy savings tool rather than solely a performance enhancement.

   For a typical urban macro site carrying 500 GB/day per cell, a 30% energy reduction translates to approximately

   2.1 MWh/day/cell in saved electrical energy, with correspond- ing CO2 emission reductions dependent on the local electricity grid carbon intensity.

7. Discussion

   The trial results conrm that the combination of ELAA hardware and AHR Turbo beamforming delivers performance improvements that exceed what can be achieved through soft- ware tuning alone on conventional 64T64R AAU platforms. The physical aperture extension from 192 to 384 elements is the enabler: it creates additional spatial degrees of freedom that the algorithm exploits but cannot generate from a smaller array.

   From a network planning perspective, the 19% cell ra-

   dius extension has signicant implications for site spacing requirements. If an operator can achieve equivalent coverage from MetaAAU with fewer macro sites or delay additional site builds the capital expenditure (CAPEX) saving can offset the incremental cost of the MetaAAU hardware upgrade relative to a conventional AAU refresh.

   The asymmetric user count gains between Site A (urban) and Site B (suburban) suggest that the capacity benet scales with the density of users within the newly-accessible coverage area, which is higher in urban than suburban environments. Operators should therefore prioritise MetaAAU deployment in dense urban coverage zones where the dual benet of

   expanded footprint and improved spatial multiplexing is most pronounced.

   Limitations of this trial include: (i) a two-site sample is insufcient for statistical condence on specic numeric KPI values; (ii) baseline AAU absolute KPI values were not disclosed in the trial report, preventing normalised throughput efciency comparisons; (iii) drive-test data distribution across coverage areas was not detailed, limiting assessment of spatial KPI variation. Broader multi-site validation across diverse propagation environments is recommended before generalising the quantitative results.

8. Conclusion

This paper has presented a eld evaluation of Huaweis 5G MetaAAU an Extremely Large Antenna Array based Active Antenna Unit in a 5G NR n78 network. The key ndings are:

• MetaAAU delivers a 3 dB UL/DL coverage improvement through its 384-element ELAA, extending cell radius by 1820% compared with a conventional 192-element 64T64R AAU.

• The AHR Turbo adaptive beamforming algorithm trans- lates the expanded aperture into 2030% cell-average and 3040% cell-edge throughput gains, with 1540% more active users per cell.

• Despite the doubled antenna count, the SDIF architecture and intelligent power management achieve approximately 30% energy savings per bit, making MetaAAU a net energy-positive upgrade.

• Both trial sites showed average TA increases of 124.8 m and 125.97 m respectively, with total trafc growing by 21.48% and 10.03% after MetaAAU activation.

MetaAAU represents a ELAA principles that bridges the gap between academic near-eld antenna research and deploy- able network infrastructure. For operators seeking to simulta- neously improve coverage, capacity, and energy efciency in mid-band 5G RAN, MetaAAU provides a compelling single- hardware-upgrade path. Future work should extend the trial to 10+ sites across diverse morphologies and quantify the CAPEX-payback period relative to new-site builds.

References

1. 3GPP, NR; Physical Channels and Modulation, 3GPP TS 38.211, Release 17, Dec. 2021.

2. E. Bjo¨rnson, J. Hoydis, and L. Sanguinetti, Massive MIMO networks: Spectral, energy, and hardware efciency, Foundations and Trends in Signal Processing, vol. 11, no. 34, pp. 154655, 2017.

3. L. Liu, C. Oestges, J. Poutanen, K. Haneda, P. Vainikainen, F. Quitin,

   F. Tufvesson, and P. D. Doncker, The COST 2100 MIMO channel model, IEEE Wireless Commun., vol. 19, no. 6, pp. 9299, 2012.

4. S. Sesia, M. Baker, and I. Touk, NR The Next Generation Wireless Access Technology. Cambridge, UK: Cambridge University Press, 2019.

5. O¨ . T. Demir, E. Bjo¨rnson, and L. Sanguinetti, Foundations of user- centric cell-free massive MIMO, Foundations and Trends in Signal Processing, vol. 14, no. 34, pp. 162472, 2021.

6. T. L. Marzetta, Noncooperative cellular wireless with unlimited num- bers of base station antennas, IEEE Trans. Wireless Commun., vol. 9, no. 11, pp. 35903600, Nov. 2010.

7. J. Hoydis, S. ten Brink, and M. Debbah, Massive MIMO in the UL/DL of cellular networks: How many antennas do we need? IEEE J. Sel. Areas Commun., vol. 31, no. 2, pp. 160171, Feb. 2013.

8. J. Zhang, E. Bjo¨rnson, M. Matthaiou, D. W. K. Ng, H. Yang, and

   D. J. Love, Prospective multiple antenna technologies for beyond 5G,

   IEEE J. Sel. Areas Commun., vol. 38, no. 8, pp. 16371660, Aug. 2020.

9. A. Pizzo, T. L. Marzetta, and L. Sanguinetti, Spatially-stationary model fr holographic MIMO small-scale fading, IEEE J. Sel. Areas Commun., vol. 38, no. 9, pp. 19641979, Sep. 2020.

10. M. Cui and L. Dai, Channel estimation for extremely large-scale MIMO: Far-eld or near-eld? IEEE Trans. Commun., vol. 70, no. 4,

    pp. 26632677, Apr. 2022.

11. X. Huang, Y. Jay Guo, A. Abbosh, and Y. Wang, Signal direct injection feed for large-scale active antenna array, IEEE Trans. Antennas Propag., vol. 70, no. 1, pp. 342351, Jan. 2022.

12. O. El Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath, Spatially sparse precoding in millimeter wave MIMO systems, IEEE Trans. Wireless Commun., vol. 13, no. 3, pp. 14991513, Mar. 2014.

13. Mohammed Yahiya Pasha Gulam 2L-4H Beams Antenna, IJERT Access, Volume 15, Issue 05 , May 2026.